Vehicle Classification on Low-resolution and Occluded images: A low-cost labeled dataset for augmentation

Video image processing of traffic camera feeds is useful for counting and classify1 ing vehicles, estimating queue length, traffic speed and also for tracking individual 2 vehicles. Even after over three decades of research, challenges remain. Vehicle 3 detection is especially challenging when vehicles are occluded which is common 4 in heterogeneous traffic. Recently Deep Learning has shown remarkable promise 5 in solving many computer vision tasks such as object recognition, detection, and 6 tracking. We explore the promise of deep learning for vehicle detection and classifi7 cation. However, training deep learning architectures require huge labeled datasets 8 which are time-consuming and expensive to acquire. We circumvent this problem 9 by data augmentation. In particular, we show that by properly augmenting an exist10 ing large general (non-traffic) dataset with a small low-resolution heterogeneous 11 traffic dataset (that we collected) we can obtain state-of-the-art vehicle detection 12 performance. This result is expected to further encourage the wide-spread use of 13 deep learning for traffic video image processing. 14